Exploring subcellular responses of prostate cancer cells to X-ray exposure by Raman mapping

Abstract

Understanding the response of cancer cells to ionising radiation is a crucial step in modern radiotherapy. Raman microspectroscopy, together with Partial Least Squares Regression (PLSR) analysis has been shown to be a powerful tool for monitoring biochemical changes of irradiated cells on the subcellular level. However, to date, the majority of Raman studies have been performed using a single spectrum per cell, giving a limited view of the total biochemical response of the cell. In the current study, Raman mapping of the whole cell area was undertaken to ensure a more comprehensive understanding of the changes induced by X-ray radiation. On the basis of the collected Raman spectral maps, PLSR models were constructed to elucidate the time-dependent evolution of chemical changes induced in cells by irradiation, and the performance of PLSR models based on whole cell averages as compared to those based on average Raman spectra of cytoplasm and nuclear region. On the other hand, prediction of X-ray doses for individual cellular components showed that cytoplasmic and nuclear regions should be analysed separately. Finally, the advantage of the mapping technique over single point measurements was verified by a comparison of the corresponding PLSR models.

Figure 1

PLSR models for cells fixed 0 h (A,B), 24 h (C,D), and 48 h (E,F) after irradiation: predicted versus applied X-ray dose (A,C,E) and regression coefficient β plots (B,D,F). Average spectra of the whole cell were used as the PLSR model input.

Figure 2

Results of a cell metabolic activity test (MTT assay) for 0 h, 24 h, and 48 h post-irradiation.

Figure 3

PLSR models for the cytoplasmic region of cells fixed 0 h (A,B), 24 h (C,D), and 48 h (E,F) after irradiation: predicted versus applied X-ray dose (A,C,E) and regression coefficient β plots (B,D,F). Average spectra of the cytoplasmic region were used as the PLSR model input.

Figure 4

PLSR models for the nuclear region of cells fixed 0 h (A,B), 24 h (C,D), and 48 h (E,F) after irradiation: predicted versus applied X-ray dose (A,C,E) and regression coefficient β plots (B,D,F). Average spectra of the nuclear region were used as the PLSR model input.

Figure 5

Selected Raman maps of predicted X-ray dose for 0 h (A,D,G), 24 h (B,E,H), and 48 h (C,F,I) calculated on the basis of PLSR models for the whole cell (from Fig. 1). Raman maps are presented for 10 Gy (A–C), 30 Gy (D–F), and 50 Gy (G–I). Bar graphs of predicted X-ray dose for 0 h (J), 24 h (K), and 48 h (L) calculated on the basis of the PLSR models for the average spectra of the whole cell (Nuc + Cyt), cytoplasm (Cyt), and the nuclear region (Nuc).

Figure 6

Bar graphs of predicted X-ray dose for 0 h (A,D,G), 24 h (B,E,H), and 48 h (C,F,I) calculated on the basis of the selected PLSR models (PLSR models for cytoplasm (A–C), nucleus (D–F), and the whole cell, cytoplasm, and the nucleus (G–I)) for the average spectra of the whole cell (Nuc + Cyt), cytoplasm (Cyt), and the nuclear region (Nuc).

Figure 7

Raman maps of a selected PC-3 cell irradiated with 10 Gy and fixed 48 h post-irradiation calculated on the basis of PLSR models for the whole cell (A–D) as well as cytoplasmic and nuclear regions separately (D–F). The maps show cytoplasmic (A,D) and nuclear (B,E) regions, and the combination of both regions (C,F).

Figure 8

Cumulative Y variance of PLSR models calculated for average spectra (red squares) and randomly selected single pixels (blue circles) of the whole cell (Nuc + Cyt), the nuclear (Nuc), and cytoplasmic (Cyt) regions.

Figure 9

PLSR coefficient plots from models for the whole cell (Nuc + Cyt; A–C), cytoplasm (Cyt; D–F), and the nuclear region (Nuc; G–I) calculated for cells fixed 48 h after irradiation. PLSR models were based on average spectra (Mean; A,D,G), randomly selected single pixels with the lowest predictive power (Min; B,E,H), and randomly selected single pixels with the highest predictive power (Max; C,F,I).